110 Transportation Research Record: Journal of the Transportation Research Board, No. 2639, 2017, pp. 110–118. http://dx.doi.org/10.3141/2639-14 Pavement maintenance and rehabilitation programming requires the consideration of conﬂicting objectives to optimize its life-cycle costs. While there are several approaches to solve multiobjective problems for pavement management systems, when user costs or environmental impacts are considered the optimal solutions are often impractical to be accepted by road agencies, given the dominating share of user costs in the total life-cycle costs. This paper presents a two-stage optimiza- tion methodology that considers maximization of pavement quality and minimization of agency costs as the objectives to be optimized at the pavement section level, while at the network level, the objectives are to minimize agency and user costs. The main goal of this approach is to provide decision makers with a range of optimal solutions from which a practically implementable one could be selected by the agency. A sensi- tivity analysis and some trade-off graphics illustrate the importance in balancing all the objectives to obtain reasonable solutions for high- way agencies. Multiobjective optimization problems at both levels are solved using genetic algorithms. The results of a case study indicate the applicability of the methodology. Decision making is a crucial component of pavement management systems (PMSs) that must be effective in optimizing the life-cycle costs of pavement networks. Generally, highway agencies aim to make the best use of their available budget to keep pavements at conditions that ensure a safe and comfort riding. A vast amount of research focused on optimizing agency expenses for a pavement life cycle while maximizing pavement quality has been done (1–4). Concerns about user costs and environmental impacts, which are a signiﬁcant parcel of pavement life-cycle costs, have been gaining more interest (5–7 ). However, when user costs (C U ) assume the same relative importance and are considered at the same decision level as agency costs (C A ), the latter have to be signiﬁcantly increased to reduce C U . Thus agencies usually ﬁnd the inclusion of C U impractical in their decision-making approach at the network level (8). Different objectives and constraints must be considered at differ- ent decision levels to ﬁnd a balanced compromise among conﬂict- ing objectives in pavement management. A bottom-up method was presented by Robelin and Madanat that solves a ﬁrst optimization problem for each facility and a second one at the system level (9). However, their approach is limited to risk-based optimization of maintenance and rehabilitation (M&R) strategies. Another meth- odology is presented in Yeo et al., where a set of activities for each facility was ﬁrst deﬁned, indicating optimal and subsequent sub- optimal alternatives (10). The objective was to ﬁnd an optimal combination of M&R activities that minimizes the total life-cycle cost of the system within the budget constraint for the current year. A decentralized multidistrict optimization was proposed by Farhan and Fwa, where the ﬁrst phase focuses on establishing the needs and funds requirement for regional administrations given performance targets, while at the second phase, budget and equity constraints are imposed at the central level (11). This paper presents a two-level methodology that intends to effec- tively consider C U in the decision-making process. M&R costs and pavement quality are considered at the pavement section level of optimization, while C U , which can be directly related to environmental impacts, are taken into account at the network level of optimization. It consists of a bottom-up approach that provides section-speciﬁc M&R alternatives that are used as inputs for networkwide decision making. Objectives are to maximize pavement quality and to minimize C U and C A , with the latter being a main objective at both optimization levels. It is expected that the proposed methodology will provide the decision maker with a range of optimal solutions from which C U minimization can be balanced with reasonable extents of C A sacriﬁce. METHODOLOGY Performance Indicators For the present work, the pavement condition is classiﬁed by perfor- mance indicators (PIs), combined performance indicators (CPIs), and a global performance indicator (GPI), as suggested by the European Cooperation in Science and Technology in its COST Action 354 report, referred to here as COST354 (12). Because there are several PIs and alternatives to determine them, the ones selected for this work are presented in Table 1. The calculation of the PI cracking (PI_CR) is made by applying the maximum weights for alligator, longitudinal, transverse, and block cracking as indicated in COST354 for ﬂexible pavements. To calcu- late the CPIs, Alternative 1 given by COST354 is used, as indicated in Equations 1 to 3: I p I I I i n CPI min 5; 100 , ,... , ( 1) 1 2 3  ( ) = +       Coordination of User and Agency Costs Using Two-Level Approach for Pavement Management Optimization André V. Moreira, Tien F. Fwa, Joel R. M. Oliveira, and Lino Costa A. V. Moreira and J. R. M. Oliveira, Center for Territory, Environment, and Construction, School of Engineering, University of Minho, Campus de Azurém, 4800-058 Guimarães, Portugal. T. F. Fwa, Department of Civil and Environ- mental Engineering, National University of Singapore, 10 Kent Ridge Crescent, Singapore 119260. L. Costa, ALGORITMI Research Center, School of Engineering, University of Minho, Campus de Gualtar, 4710-057 Braga, Portugal. Corresponding author: A. V. Moreira, avm@civil.uminho.pt.